Configuration

SimCLR histology encoder + random forest (HEST; Ciga et al.)

Published histology patch encoder whose frozen features the HEST authors scored with a Random Forest regression head.

Identified as SimCLR histology encoder of Ciga et al.. The source table prints Ciga.

Configuration: Frozen patch embeddings scored by HEST with its 70-tree random-forest regression head.

HEST §5.2 lists 'Ciga Ciga et al. [2022] (SimCLR pretrained on public histology data)' among its ten patch encoders, and Table 1 scores their embeddings with a 70-tree random forest. The cited paper applies SimCLR to 57 unlabelled histopathology datasets. The printed label 'Ciga' is the first author's surname, so the display name describes the method and keeps the attribution.

Regression head: Random forest with 70 trees (sklearn) mapping patch embeddings to log1p-normalised expression of the top 50 highly variable genes, scored by Pearson correlation.
Pretraining: SimCLR self-supervised pretraining on public histology data.

Not established: Backbone. HEST App. C.3 (p.17) and Table A12 (p.26) say ResNet-50. The original repository example loads ResNet-18 and the lab repository offers ResNet-34, 50 and 101. Which released weights HEST used is not stated. Training recipe. HEST Table A12 labels the recipe 'Supervised', which contradicts HEST §5.2, App. C.3 and the original paper. Checkpoint. No checkpoint file or revision is identified.

HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis; Self supervised contrastive learning for digital histopathology (arXiv:2011.13971v2); Self-supervised histopathology repository README (ozanciga/self-supervised-histopathology); Self-supervised histopathology repository README (martellab-sri/self-supervised-histopathology) · HEST arXiv:2406.16192v1 §5.2, p.6; Table 1, p.7; App. C.3, pp.14 and 17; Table A12, p.26. Ciga et al. arXiv:2011.13971v2 abstract, p.1. github.com/ozanciga/self-supervised-histopathology README at f7a93798911c861efa297fd9c9b9b1f9ece0c1c0; github.com/martellab-sri/self-supervised-histopathology README at 22d3440f8d94d6e367304f1e50b82e9ea930dbd7.. AI-assisted source review on 2026-09-24; no human scientific review.

10 evaluations · 10 results

Overview

An explanatory profile has not yet been reviewed for this record.

Consult the linked sources for architecture or protocol details. Missing evidence is not evidence of a missing capability.

Evaluations and results

10 evaluations · 10 results. Different protocols are not a single leaderboard.

Sorted by Pearson (higher is better). The best value in each column is highlighted. Decimals are rounded for display; each value links to the printed value and its source.

Tested configuration	Protocol and dataset	Pearson
SimCLR histology encoder + random forest (HEST; Ciga et al.)Evaluation	HEST-Benchmark LUNG: Gene expression prediction from histology, Lung, HEST-Benchmark LUNG (HEST-Benchmark split)	0.515 ±0.02 (best in column)
SimCLR histology encoder + random forest (HEST; Ciga et al.)Evaluation	HEST-Benchmark SKCM: Gene expression prediction from histology, Skin cutaneous melanoma, HEST-Benchmark SKCM (HEST-Benchmark split)	0.484 ±0.01
SimCLR histology encoder + random forest (HEST; Ciga et al.)Evaluation	HEST-Benchmark IDC: Gene expression prediction from histology, Invasive ductal carcinoma, HEST-Benchmark IDC (HEST-Benchmark split)	0.406 ±0.02
SimCLR histology encoder + random forest (HEST; Ciga et al.)Evaluation	HEST-Benchmark PAAD: Gene expression prediction from histology, Pancreatic adenocarcinoma, HEST-Benchmark PAAD (HEST-Benchmark split)	0.397 ±0.07
SimCLR histology encoder + random forest (HEST; Ciga et al.)Evaluation	HEST-Benchmark PRAD: Gene expression prediction from histology, Prostate adenocarcinoma, HEST-Benchmark PRAD (HEST-Benchmark split)	0.332 ±0.00
SimCLR histology encoder + random forest (HEST; Ciga et al.)Evaluation	HEST-Benchmark LYMPH_IDC: Gene expression prediction from histology, Lymph node metastasis of invasive ductal carcinoma, HEST-Benchmark LYMPH_IDC (HEST-Benchmark split)	0.218 ±0.07
SimCLR histology encoder + random forest (HEST; Ciga et al.)Evaluation	HEST-Benchmark CCRCC: Gene expression prediction from histology, Clear cell renal cell carcinoma, HEST-Benchmark CCRCC (HEST-Benchmark split)	0.127 ±0.04
SimCLR histology encoder + random forest (HEST; Ciga et al.)Evaluation	HEST-Benchmark COAD: Gene expression prediction from histology, Colon adenocarcinoma, HEST-Benchmark COAD (HEST-Benchmark split)	0.102 ±0.04
SimCLR histology encoder + random forest (HEST; Ciga et al.)Evaluation	HEST-Benchmark READ: Gene expression prediction from histology, Rectum adenocarcinoma, HEST-Benchmark READ (HEST-Benchmark split)	0.046 ±0.09
SimCLR histology encoder + random forest (HEST; Ciga et al.)Evaluation	HEST-Benchmark HCC: Gene expression prediction from histology, Hepatocellular carcinoma, HEST-Benchmark HCC (HEST-Benchmark split)	0.045 ±0.00

All 10 result rows with coverage, uncertainty and sources

Filter evaluations

Tested model or methodProtocolDatasetEvaluation setupMetricEvidence origin

Applied filters: All linked evaluations

Exact evaluated configurations and original reported results
Tested configuration	Protocol and dataset	Finding	Evidence and details
Configuration: SimCLR histology encoder + random forest (HEST; Ciga et al.)	Task: HEST-Benchmark CCRCC: Gene expression prediction from histology, Clear cell renal cell carcinoma Dataset subset: HEST-Benchmark CCRCC (HEST-Benchmark split)	0.127 ±0.04 Pearson unitless · higher Uncertainty: SD 0.04 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source SimCLR histology encoder + random forest (HEST; Ciga et al.) on HEST-Benchmark CCRCC: Gene expression prediction from histology, Clear cell renal cell carcinoma Random Forest regression with 70 trees over frozen patch features, averaged over folds or patients. Aggregation: Not reported HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis · Table 1, row(CCRCC), column(Ciga)
Configuration: SimCLR histology encoder + random forest (HEST; Ciga et al.)	Task: HEST-Benchmark COAD: Gene expression prediction from histology, Colon adenocarcinoma Dataset subset: HEST-Benchmark COAD (HEST-Benchmark split)	0.102 ±0.04 Pearson unitless · higher Uncertainty: SD 0.04 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source SimCLR histology encoder + random forest (HEST; Ciga et al.) on HEST-Benchmark COAD: Gene expression prediction from histology, Colon adenocarcinoma Random Forest regression with 70 trees over frozen patch features, averaged over folds or patients. Aggregation: Not reported HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis · Table 1, row(COAD), column(Ciga)
Configuration: SimCLR histology encoder + random forest (HEST; Ciga et al.)	Task: HEST-Benchmark HCC: Gene expression prediction from histology, Hepatocellular carcinoma Dataset subset: HEST-Benchmark HCC (HEST-Benchmark split)	0.045 ±0.00 Pearson unitless · higher Uncertainty: SD 0.00 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source SimCLR histology encoder + random forest (HEST; Ciga et al.) on HEST-Benchmark HCC: Gene expression prediction from histology, Hepatocellular carcinoma Random Forest regression with 70 trees over frozen patch features, averaged over folds or patients. Aggregation: Not reported HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis · Table 1, row(HCC), column(Ciga)
Configuration: SimCLR histology encoder + random forest (HEST; Ciga et al.)	Task: HEST-Benchmark IDC: Gene expression prediction from histology, Invasive ductal carcinoma Dataset subset: HEST-Benchmark IDC (HEST-Benchmark split)	0.406 ±0.02 Pearson unitless · higher Uncertainty: SD 0.02 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source SimCLR histology encoder + random forest (HEST; Ciga et al.) on HEST-Benchmark IDC: Gene expression prediction from histology, Invasive ductal carcinoma Random Forest regression with 70 trees over frozen patch features, averaged over folds or patients. Aggregation: Not reported HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis · Table 1, row(IDC), column(Ciga)
Configuration: SimCLR histology encoder + random forest (HEST; Ciga et al.)	Task: HEST-Benchmark LUNG: Gene expression prediction from histology, Lung Dataset subset: HEST-Benchmark LUNG (HEST-Benchmark split)	0.515 ±0.02 Pearson unitless · higher Uncertainty: SD 0.02 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source SimCLR histology encoder + random forest (HEST; Ciga et al.) on HEST-Benchmark LUNG: Gene expression prediction from histology, Lung Random Forest regression with 70 trees over frozen patch features, averaged over folds or patients. Aggregation: Not reported HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis · Table 1, row(LUNG), column(Ciga)
Configuration: SimCLR histology encoder + random forest (HEST; Ciga et al.)	Task: HEST-Benchmark LYMPH_IDC: Gene expression prediction from histology, Lymph node metastasis of invasive ductal carcinoma Dataset subset: HEST-Benchmark LYMPH_IDC (HEST-Benchmark split)	0.218 ±0.07 Pearson unitless · higher Uncertainty: SD 0.07 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source SimCLR histology encoder + random forest (HEST; Ciga et al.) on HEST-Benchmark LYMPH_IDC: Gene expression prediction from histology, Lymph node metastasis of invasive ductal carcinoma Random Forest regression with 70 trees over frozen patch features, averaged over folds or patients. Aggregation: Not reported HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis · Table 1, row(LYMPH_IDC), column(Ciga)
Configuration: SimCLR histology encoder + random forest (HEST; Ciga et al.)	Task: HEST-Benchmark PAAD: Gene expression prediction from histology, Pancreatic adenocarcinoma Dataset subset: HEST-Benchmark PAAD (HEST-Benchmark split)	0.397 ±0.07 Pearson unitless · higher Uncertainty: SD 0.07 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source SimCLR histology encoder + random forest (HEST; Ciga et al.) on HEST-Benchmark PAAD: Gene expression prediction from histology, Pancreatic adenocarcinoma Random Forest regression with 70 trees over frozen patch features, averaged over folds or patients. Aggregation: Not reported HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis · Table 1, row(PAAD), column(Ciga)
Configuration: SimCLR histology encoder + random forest (HEST; Ciga et al.)	Task: HEST-Benchmark PRAD: Gene expression prediction from histology, Prostate adenocarcinoma Dataset subset: HEST-Benchmark PRAD (HEST-Benchmark split)	0.332 ±0.00 Pearson unitless · higher Uncertainty: SD 0.00 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source SimCLR histology encoder + random forest (HEST; Ciga et al.) on HEST-Benchmark PRAD: Gene expression prediction from histology, Prostate adenocarcinoma Random Forest regression with 70 trees over frozen patch features, averaged over folds or patients. Aggregation: Not reported HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis · Table 1, row(PRAD), column(Ciga)
Configuration: SimCLR histology encoder + random forest (HEST; Ciga et al.)	Task: HEST-Benchmark READ: Gene expression prediction from histology, Rectum adenocarcinoma Dataset subset: HEST-Benchmark READ (HEST-Benchmark split)	0.046 ±0.09 Pearson unitless · higher Uncertainty: SD 0.09 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source SimCLR histology encoder + random forest (HEST; Ciga et al.) on HEST-Benchmark READ: Gene expression prediction from histology, Rectum adenocarcinoma Random Forest regression with 70 trees over frozen patch features, averaged over folds or patients. Aggregation: Not reported HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis · Table 1, row(READ), column(Ciga)
Configuration: SimCLR histology encoder + random forest (HEST; Ciga et al.)	Task: HEST-Benchmark SKCM: Gene expression prediction from histology, Skin cutaneous melanoma Dataset subset: HEST-Benchmark SKCM (HEST-Benchmark split)	0.484 ±0.01 Pearson unitless · higher Uncertainty: SD 0.01 Coverage: Not reported scored / Not reported eligible	Author-reported evaluation · Source checked Methods, coverage and source SimCLR histology encoder + random forest (HEST; Ciga et al.) on HEST-Benchmark SKCM: Gene expression prediction from histology, Skin cutaneous melanoma Random Forest regression with 70 trees over frozen patch features, averaged over folds or patients. Aggregation: Not reported HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis · Table 1, row(SKCM), column(Ciga)

Source checking is not independent reproduction. Release 2026-11-02-a7f03c9e55a5.

Use this configuration

How it works, versions and access

Additional methodological details, versions and access terms have not yet been extracted into a structured profile for this configuration. Consult the linked sources and evaluation records below.

Evidence

Source checking verifies the cited claim or transcription. It does not establish independent reproduction.

Evidence table

Inspect claims, sources and review details

Trace each statement to its source and review. A context-only reference supports the record generally; it does not verify an individual field. Source checking does not reproduce an experiment.

One row per statement and cited source. Multiple citations are not independent evaluations. Shared locators are labelled explicitly.

0 evidence rows matching the loaded filters

Claims, original sources and review scope · Release 2026-11-02-a7f03c9e55a5
Property and statement	Original source and location	Review and provenance

No evidence rows match these filters. Choose another scope or clear the search.

How to read the evidence

Sources and history

Release 2026-11-02-a7f03c9e55a5 · Record review: source checked

4 source records and release history

HEST-1k: A Dataset for Spatial Transcriptomics and Histology Image Analysis · Original source · Version pinned by URL and artifact SHA256 when available
Self supervised contrastive learning for digital histopathology (arXiv:2011.13971v2) · Original source · 2011.13971v2
Self-supervised histopathology repository README (ozanciga/self-supervised-histopathology) · Original source · f7a93798911c861efa297fd9c9b9b1f9ece0c1c0
Self-supervised histopathology repository README (martellab-sri/self-supervised-histopathology) · Original source · 22d3440f8d94d6e367304f1e50b82e9ea930dbd7

Download this release (gzip)

Technical metadata and extraction receipts

Stable ID: hest-method-ciga

areas: cells-tissues
source locator: Table 1, column(Ciga)
missing metadata: checkpoint revision: reason: unreported; parameters: reason: unextracted
source label: Ciga
source identity: status: resolved; label form: author_surname; display name: SimCLR histology encoder + random forest (HEST; Ciga et al.); identity: SimCLR histology encoder of Ciga et al.; configuration: Frozen patch embeddings scored by HEST with its 70-tree random-forest regression head.; basis: HEST §5.2 lists 'Ciga Ciga et al. [2022] (SimCLR pretrained on public histology data)' among its ten patch encoders, and Table 1 scores their embeddings with a 70-tree random forest. The cited paper applies SimCLR to 57 unlabelled histopathology datasets. The printed label 'Ciga' is the first author's surname, so the display name describes the method and keeps the attribution.; source ids: evidence-expansion-p2-hest-cached-636099a73dee; source-label-ciga-arxiv-3b08dd7e; source-label-ciga-readme-f7a93798; source-label-martellab-readme-22d3440f; source locator: HEST arXiv:2406.16192v1 §5.2, p.6; Table 1, p.7; App. C.3, pp.14 and 17; Table A12, p.26. Ciga et al. arXiv:2011.13971v2 abstract, p.1. github.com/ozanciga/self-supervised-histopathology README at f7a93798911c861efa297fd9c9b9b1f9ece0c1c0; github.com/martellab-sri/self-supervised-histopathology README at 22d3440f8d94d6e367304f1e50b82e9ea930dbd7.; known details: label: Regression head; value: Random forest with 70 trees (sklearn) mapping patch embeddings to log1p-normalised expression of the top 50 highly variable genes, scored by Pearson correlation.; source ids: evidence-expansion-p2-hest-cached-636099a73dee; source locator: HEST §5.2, p.6; App. C.3, p.14; label: Pretraining; value: SimCLR self-supervised pretraining on public histology data.; source ids: evidence-expansion-p2-hest-cached-636099a73dee; source-label-ciga-arxiv-3b08dd7e; source locator: HEST §5.2, p.6; Ciga et al. abstract, p.1; unknown: label: Backbone; note: HEST App. C.3 (p.17) and Table A12 (p.26) say ResNet-50. The original repository example loads ResNet-18 and the lab repository offers ResNet-34, 50 and 101. Which released weights HEST used is not stated.; label: Training recipe; note: HEST Table A12 labels the recipe 'Supervised', which contradicts HEST §5.2, App. C.3 and the original paper.; label: Checkpoint; note: No checkpoint file or revision is identified.; original name: Ciga; original description: Published histology patch encoder whose frozen features the HEST authors scored with a Random Forest regression head.; review: method: automated_source_review; date: 2026-09-24; note: AI-assisted review against the cited primary sources. No human scientific review. Values, locators and comparison conditions are unchanged.

Related records

Suggest a correction